Research Agent – Displayr Help

Making it easier for users to find and share data stories is Displayr's bread and butter, and with the new Research Agent feature, users can use AI to actually analyze and create the report! Think of Research Agent as a data analysis intern; you provide the data and guidance on what questions and information you want covered in the report, and Research Agent automatically analyzes the data and creates the report for your review. It will automatically find and highlight key insights found in your data, as well as any specific questions you need investigated. This guide explains how to use it effectively and get the most out of its capabilities. It contains the following sections:

How to Use Research Agent
Frequently Asked Questions
Best Practices for Using Research Agent

Requirements

Displayr AI needs to be enabled. See Opting In and Out of Displayr AI.
It is recommended you either:
- Clean and tidy your data using the built-in prompt. See How to Automatically Check and Prepare Your Data with Data Preparation Agent for details.
- Clean and tidy your data manually, see How to Check and Clean Data After Importing, before running Research Agent.

How to Use Research Agent

Create a Displayr document. See How to Create a New Document in Displayr.
Add a data set. See How to Import Data Into Displayr.
Clean and tidy your data. Research Agent works best when the data is structured, consistent, and free of errors. You can either:
- Clean and tidy your data manually, see How to Check and Clean Data After Importing.
- Or use the Data Preparation Agent within Research Agent as described below. For more details on what the Data Preparation Agent does, see How to Automatically Check and Prepare Your Data with Data Preparation Agent.
Set up weights or filters to apply to the data for the report.
Research Agent will apply all filters where the object inspector > General > Comment has the text Research Agent.
Research Agent can only use one weight (which must have Usable as a weight checked in its object inspector). If there is only one weight in the data set, that will be used; otherwise, add Research Agent to the object inspector > General > Comment of the weight variable you wish to use.
In the Report tree, click + Add to report, or click and select Research Agent.
If you have more than one data set in your document, you'll be asked to Select your data set to use. Select it and click OK.
If you want to Prepare your data automatically using the Data Preparation Agent, select your data checks. Otherwise, to use the data as-is in your report, click Skip and go to Step 8.
See our technical reference, Data Preparation Agent, for a description of each check.
Click Continue, and you will see a working animation while the AI prepares your data. This process reviews data for every variable in your data set, and may take several minutes for large data sets.
Next, the What do you want to know? window pops up, with a suggested prompt on how to create the report. You can modify this prompt or create your own from scratch if you want. See Best Practices for Using Research Agent below.
Using the suggested prompt:
Fill in the Sample Description and Background sections of the prompt to provide more context to the AI to use during analysis.
Review the list of Research Questions and modify as needed.
Select the data to be analyzed and click Continue. You will see a working animation while the AI prepares your data. This process reviews data for every variable in your data set, and may take several minutes for large data sets.
Review and refine the report.

You can manually modify any of the report.

If you ran the Data Preparation Agent, a new page named Data Preparation will be in the report outlining the changes for each check. You can review these and modify/delete any variables in the Data Sources tree as needed.

Automatically update commentary on a page by right-clicking it in the Report tree and selecting Add Commentary.

To update the summary page(s) at the beginning, delete the existing summary page(s) via right-click > Delete, select all the pages you wish to be summarized, right-click, and select Add Commentary.

Remember: AI makes mistakes. It's important to check and correct the output. This typically involves:
1. Modifying the tables behind the visualizations (e.g., applying filters, weights, or cleaning).
2. Adjusting the visualizations.
3. Editing the automatically generated commentary.
OPTIONAL: Create a dynamically updating dashboard.

Create filters. See Filtering.

Consider deleting the executive summary pages. This will not update automatically.

In the Report tree, select all the pages you wish to update, right-click, and select Auto-Updating Interpretation.

Frequently Asked Questions

Why is it labeled as Beta?

We're still ironing out some kinks and looking for feedback. Contact support@displayr.com with feedback and questions.

Why doesn't the report update automatically when data changes?

To preserve your manual edits. If the report regenerated automatically, you could lose any customizations you've made. To generate new interpretations, select all the pages you wish to update, right-click, and select Add Commentary or Auto-Updating Interpretation.

Why can’t I select all my data?

To maintain quality, Research Agent excludes:

Large tables, where it seems unlikely they can be useful (e.g., crosstabs with more than 1,000 cells).
Ranking and Experiment variable sets.

These are often problematic, and including them can degrade the analysis and use up your AI allowance unnecessarily.

You can work around this limitation by manually creating a table and using Add Commentary.

What does the Data Preparation Agent do?

Data Preparation Agent (the Prepare your data prompt) is built into Research Agent and saves you time cleaning and creating new variables. Read more about the functions that it runs in How to Automatically Check and Prepare Your Data with Data Preparation Agent.

Does Research Agent ignore NET columns in crosstabs?

Yes. Since NET values summarize what's already in the data, including them would lead to repetitive conclusions. Research Agent skips these to avoid redundancy.

Can I modify the report?

Yes. Every aspect can be edited/modified.

Can I modify the prompt?

Yes. The What do you want to know form is your prompt. It's saved and can also be modified on the Background page created in the report.

Does it work in languages other than English?

Yes. It will automatically work in most languages. To control the language of the output, add this information to the What do you want to know prompt.

Can I control the defaults for charts and text?

Yes. Text defaults and default color settings are controlled via the page master.

Research Agent has 10 alternative ways of displaying the data, based on the properties of the data. The 10 defaults are shown below. To change any of these, create a visualization as per your requirements and save it as a template, giving it the names as shown below (e.g., to replace the pie chart with something else, create a visualization and save it as Research Agent Template Six or Fewer Mutually Exclusive Categories).

Best Practices for Using Research Agent

Clean and tidy your data.
The better the input, the better the output. The Research Agent comes with data cleaning capabilities via the Data Preparation Agent. It will perform a multitude of checks on your data and clean your variable sets appropriately. Additionally, we recommend using Text Categorization for open-ended questions. There is a built-in function as part of the Data Preparation Agent. For spontaneous awareness data, you can use List of Items instead, if you'd like. See Get Started With Text Analytics for guidance. If you prefer to clean and tidy the data manually, see How to Clean and Tidy Data for more methods to improve your data input.
Provide context.
When answering Research Agent’s questions, be specific. Mention:
- Sample profile: Who was screened out? What’s the population?
- What decisions are you trying to make?
Split large studies.
For complex or large data sets, it's often better to run Research Agent separately on different sections of the data.
Always review and correct.
Research Agent is powered by AI - it’s fast, but not perfect. Human oversight is essential.